Australasian Marketing Journal 20 (2012) 282–296
Contents lists available at SciVerse ScienceDirect
Australasian Marketing Journal
journal homepage: www.elsevier.com/locate/amj
A critical review of classical conditioning effects on consumer behavior
Chanthika Pornpitakpan
University of Macau, Taipa, Macau, China
a r t i c l e
i n f o
Article history:
Received 15 July 2011
Revised 14 June 2012
Accepted 3 July 2012
Available online 28 July 2012
Keywords:
Classical conditioning
Pavlovian conditioning
Literature review
a b s t r a c t
This paper reviews extant research in classical conditioning effects in consumer behavior and advertising
contexts to determine whether they are real or illusory. The empirical results reveal that in cases where
classical conditioning effects were found, they could be countermined by the deficiencies in research
methodologies, demand artifacts, the mediating role of contingency awareness, or some alternative
mechanisms. In cases where the effects were not observed, the failure could be attributed to violations
of the conditions for classical conditioning to occur or absence of contingency and demand awareness.
It is concluded that thus far there has been no convincing evidence for classical conditioning effects on
consumer behavior. Suggestions for future research in this area are presented.
Ó 2012 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.
1. Introduction
Inspired by classical conditioning principles, many ads show the
advertised product together with celebrities or pleasant stimuli (objects, scenes, persons, and so forth) once or several times with a
hope that positive feelings from those stimuli will transfer to the
product and thus inducing its liking. Classical conditioning has been
generally accepted in consumer behavior literature as a mechanism
producing advertising effects (Schiffman and Kanuk, 2010), as a possible mechanism in the peripheral route of persuasion (Edell and
Burke, 1984; Petty et al., 1983), and as pertinent in passive consumption context (Gorn, 1982; Greenwald and Leavitt, 1984).
According to the classical conditioning model of learning, which
is based on Pavlov’s (1927) work, an unconditioned stimulus (hereafter referred to as US or USs for the plural form) is a biologically
significant stimulus such as food, pain, electric shock that generates a response (for example, salivation when seeing certain foods)
from the start; this response is referred to as an unconditioned response. Repeated pairings of a conditioned stimulus (hereafter referred to as CS or CSs for the plural form, for example, the ring of a
bell) with an US (for example, meat paste) will enable the CS to elicit a conditioned response (for example, salivation) in an unconscious and automatic manner. When the US is an affect (Razran,
1938), for instance, music and humor, the conditioning may be referred to as affective conditioning.
1.1. Types of classical conditioning
Within the paradigm of classical conditioning, it has been proposed that a distinction be made between different types of conditioning, namely, signal learning and evaluative learning (Baeyens
E-mail addresses: ynvynv@yahoo.com, ynvynv@gmail.com
and De Houwer, 1995; Baeyens et al., 1998; Hammerl and Grabitz,
1996). In the signal-learning notion of classical conditioning, an
organism engages a higher cognitive process and learns the ‘‘ifthen’’ relationship between the CS and the US (Rescorla and
Wagner, 1972). Repeated pairings of the two stimuli to be associated is essential in establishing and strengthening their associative
link (Martindale, 1991). Through a signal learning process
(Rescorla, 1988), increased repetition of the pairing of two stimuli
fortifies confidence that the presence of one stimulus predicts the
presence of the other. The contingency or statistical correlation
between the CS and the US is an important determinant of signal
learning.
Evaluative (attitude) conditioning, on the other hand, concerns
the acquisition of preferences and refers to the change in valence of
initially neutral CSs after pairing with positive or negative USs.
Evaluative conditioning is usually conceptualized as a form of evaluative learning that occurs without awareness of the CS–US contingencies (De Houwer et al., 2001; Stahl et al., 2009). In a
typical evaluative conditioning study (e.g., De Houwer et al.,
2001; Walther, 2002), a subjectively neutral stimulus is repeatedly
paired with a subjectively liked or disliked stimulus, leading to a
valence shift in the formerly neutral stimulus. That is, the CS in
an evaluative conditioning paradigm does not attain a predictive
value but simply obtains the affective qualities of the US.
Three major characteristics of evaluative conditioning are as
follows. First, evaluative conditioning does not seem to depend
on contingency awareness of the CS and the US (Baeyens et al.,
1990; De Houwer et al., 2001). Second, it does not appear to rely
on the statistical CS–US contingency but seems to be sensitive to
contiguity, that is, to spatiotemporal CS–US co-occurrences
(Baeyens et al., 1993; De Houwer et al., 2001). Therefore, weak
contingency in an evaluative learning paradigm (e.g., single CS or
US presentations in the acquisition phase) does not automatically
1441-3582/$ - see front matter Ó 2012 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ausmj.2012.07.002
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
reduce conditioning, as would be the case in signal learning (Baeyens et al., 1993). Third, after successful evaluative conditioning,
single CS presentations may not alter its valence; in other words,
evaluative conditioning seems to be extinction resistant (Baeyens
et al., 1988; De Houwer et al., 2001).
However, it has been criticized that the conclusions about evaluative conditioning without CS–US contingency awareness often
relied on questionable experimental designs or failed to capture
subtle but substantial manifestations of such awareness (Field,
2000; Field and Davey, 1999; Hammerl, 2000; Lovibond and
Shanks, 2002). This view/critique implies that the differentiation
between signal learning and evaluative conditioning as two types
of classical conditioning is rather obscure and in fact, evaluative
conditioning is merely a situation when CS–US contingency awareness cannot be detected and the focus is on the change in valence
of initially neutral CSs after pairing with positive or negative USs. It
follows from this view/critique that CS–US contingency awareness
underlies any observed classical conditioning effect.
1.2. Objectives and contributions of the study
Consumer researchers have empirically investigated classical
conditioning effects after Gorn’s (1982) experiments. The results,
unfortunately, have been mixed. Given three decades of classical
conditioning studies in consumer behavior and the prevalent
application of classical conditioning principle knowingly or
unknowingly in advertising, this study aims to review classical
conditioning research in the realm of consumer behavior and
advertising to find out whether the effects are real or illusory. In
addition, the study discusses the common weaknesses of research
in this area and suggests what future research in this field should
improve. Sharing the same view discussed in the preceding paragraph, this review covers studies in both signal learning and evaluative conditioning paradigms of classical conditioning as long as
they involve consumer behavior and advertising. Due to space constraint, not all studies reviewed will be evaluated in detail.
The contributions of this study are two. First, in terms of academic contribution, this study pinpoints common weaknesses of
research in classical conditioning and suggests ways to improve research in this area in order to increase internal validity and yield
results that are more credible.
Second, in terms of managerial contribution, this study provides
guidelines for advertisers whether they should use the classical
conditioning principle in designing and producing ads, which are
usually costly. For example, a television commercial involving
non-celebrity actors, storyboarding, and script writing with a
length of 30-s costs on average US$3500–US$35000 to produce
(Maus Media Group, 2011). The cost to create a full page colored
ad by a freelancer may start from US$2500 and much more by a
large advertising agency. The media expenses associated with
showing the ads are even much higher. For instance, the average
price of a 30-s television spot in the first quarter of 2011 in the
USA was US$108,956 (Crupi, 2011). A full-page ad in Cosmopolitan
magazine costs US$244,100 for colored and US$195,300 for blackand-white ads, respectively (Cosmopolitan, 2012). Clearly, it is
essential that marketers know whether ads based on classical conditioning concepts are effective given such large expenditure of
producing and displaying ads.
283
2.1. The CS lacking predictiveness of the US
In order for classical conditioning to occur, the CS should predict the US. The CS will have low predictiveness of the US if (i) both
stimuli are presented simultaneously; (ii) the CS is presented constantly, followed occasionally by the US (Brown and Jenkins,
1968); (iii) either the US or the CS is frequently encountered alone;
(iv) the US comes before the CS (i.e., backward conditioning); and
(v) the CS and the US are presented randomly with respect to each
other. When this predictive relation is violated, the conditioned
behavior will not be established.
However, Shurtleff and Ayres (1981) found backward conditioning. Spetch et al. (1981) reviewed experimental evidence and
concluded that backward conditioning could produce effects similar to those obtained from forward conditioning. In addition, such
effects could not be explained by factors other than stimulus pairing. Rachman’s (1991) review suggests that conditioning can occur
even when the CS and the US are not contiguous.
In order to say that conditioning occurs in an experiment,
appropriate control groups are needed for comparison with the
conditioning groups (Rescorla, 1967). The experimental group is
exposed to the CS, followed by the US. A random control group is
exposed to the same quantity of CSs and USs as the experimental
group, but these stimuli are presented randomly with respect to
each other. Classical conditioning can be said to occur only if there
is a response in the experimental group but not in the random control group. This way, the differences between the two groups can
be explained by the degree to which the CS predicts the US and
cannot be attributed to differences in familiarity with the CS or
the US, i.e., the mere exposure effect (Zajonc, 1968) or to any interaction between them (pseudoconditioned responses).
Even though the CS precedes and predicts the US, conditioning
may not occur under the situations detailed below.
2.2. Overshadowing
Overshadowing can prevent classical conditioning from occurring (Pavlov, 1927). In an overshadowing procedure, two CSs different in salience, such as a loud noise and a dim light, are
presented together before the US. Conditioning may occur to the
more salient CS (in this example, the loud noise) only because
the less salient CS is overshadowed by the more salient one. This
suggests that salient stimuli coinciding with the target CS (for instance, a highly sexy model and a brand in an ad) be removed.
2.3. Blocking
Blocking can forbid classical conditioning (Kamin, 1969). In a
blocking procedure, an individual is given experience that CS1 is
predictive of an US (for example, dark clouds predict rain). Later,
CS1 and CS2 (for example, a barometer that also predicts rain)
are presented together, followed by the same US. No conditioned
response occurs to CS2, however. In effect, prior experience with
CS1 blocks conditioning to CS2. In marketing, this indicates that
a familiar US should not be used. For instance, using a celebrity endorser who has been well established as an endorser for other
products suppresses forming an association between that celebrity
and a newly endorsed product (Till, 1998).
2. Obstacles and methodological requirements for classical
conditioning
2.4. US pre-exposure effect
It is essential that the conditions hindering classical conditioning be examined so that conflicting studies can be assessed. Primarily based on McSweeney and Bierley’s (1984) review, the
obstructions to classical conditioning are as follows.
Classical conditioning will not occur if individuals have encountered the US alone (Mis and Moore, 1973; Rescorla, 1973). The
implication is the same as that for the blocking effect, namely, a
familiar US, such as a famous song, should not be used.
284
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
2.5. Latent inhibition
Latent inhibition also forestalls classical conditioning (Lubow,
1973). In a latent inhibition procedure, the CS is presented alone
several times before it is paired with the US, resulting in little conditioning. A marketing implication of latent inhibition is that using
a familiar CS, such as a well-known product, will not be effective,
and it is easier to classically condition responses to unfamiliar
CSs such as new brands.
2.6. Garcia effect
Garcia effect (Garcia and Koelling, 1966) refers to an improper
match between the CS and the US. It cannot be assumed, as a result, that any stimulus can be used as a CS or an US. For example,
in Garcia and Koelling’s experiments, an aversion (conditioned response) was easily conditioned to a flavor (CS1) that was followed
by sickness (US1) but not by shock (US2). In other words, the flavor-sickness CS–US pairing was appropriate but the flavor-shock
pairing was not. Similarly, an aversion (conditioned response)
could develop to a light and a noise (CS2) that were followed by
shock (US2) but not by sickness (US1). That is, the light/noiseshock CS–US pairing was suitable but the light/noise-sickness pairing was not.
It has been speculated that stimuli that biologically belong together, such as tastes and sickness, results in effective conditioning. Generally, it is easier to build an associative link between
two stimuli when they are similar or related to each other (McSweeney and Bierley, 1984). In marketing, the choice of the US
should correspond to the associations the brand already has or
seeks to have. For instance, Kim et al. (1996) demonstrated that
the pairing of pizza delivery (CS) with a race car (US) improved
respondents’ beliefs about the speed of the pizza delivery service.
‘‘Starter’’ athletic apparel employed the durable and venerable
baseball player Cal Ripken to emphasize that Starter jackets were
also durable. In the 1990’s, Pepsi used many endorsers such as Michael J. Fox, Madonna, and Michael Jackson to enhance its youth
image. Authoritative, genteel, and cultured John Houseman fit well
with the investment firm Smith Barney. However, McDonald’s
used John Houseman for only a short time because his image did
not go well with a family-oriented fast food restaurant (Till, 1998).
2.7. Insufficient pairings
The acquisition speed of conditioned responses can range from
one to many pairings of the CS and the US. The acquisition of conditioned responses will be faster for CSs that are more salient
(Kamin and Brimer, 1963; Rescorla, 1972), for USs that are stronger
(Pavlov, 1927; Wagner et al., 1964), and for longer inter-trial intervals (Terrace et al., 1975), which refers to longer time between successive CS–US pairings. Salient CSs are those that are intense
physically (for example, brighter lights) or have acquired some
psychological importance to a person (for example, his or her
own name) (McSweeney and Bierley, 1984).
3. Evidence for classical conditioning effects
The evaluation of most studies below is based on the above
impediments and methodological requirements for classical conditioning. The studies can be grouped by the following methodological scheme: 2 (familiarity of the CSs: familiar, unfamiliar) by 2
(familiarity of the USs: familiar, unfamiliar) by 2 (number of repeated CS–US pairings: single, multiple) by 3 (order of CSs and
USs in the pairing: CSs preceding USs, simultaneous, USs preceding
CSs). The full classification scheme results in 24 groups in total,
which seems to be unnecessarily complicated. Therefore, the three
patterns of ‘‘order of CSs and USs in the pairing’’ are merged but
will be mentioned for each study. Some groups contain no studies
and therefore do not appear as sub-section headings.
3.1. Studies using familiar CSs, familiar USs, and a single CS–US pairing
Gorn (1982) provided the first support for classical conditioning
in consumer behavior. He showed students slides of either a beige
or a blue pen (a CS) while having them listen to either liked or disliked music (an US). When given a choice of taking a beige or a blue
pen as a gift, participants chose the pen associated with the liked
music and avoided the one associated with the disliked music.
Gorn’s (1982) findings generated some concerns that inspired
many later studies to replicate, extend, or refute his findings. It
seemed rather difficult for a single simultaneous pairing of the
CS and the US to produce the conditioned response because the
CS will not predict the US if they are presented simultaneously.
While classical conditioning may occur after one trial (Shurtleff
and Ayres, 1981), it usually entails a very strong US such as an intense shock or a nauseating drug (Bierley et al., 1985). In addition,
participants in the liked music condition (one-minute extract of
music from the movie ‘‘Grease’’) were likely to be familiar with
the US (the music), constituting an US pre-exposure effect or even
a blocking effect if it had been elsewhere constantly associated
with some other CSs. Although no evidence in the study indicated
that participants had heard the music with other stimuli, since the
music was from a popular movie, the chance of participants not
having heard this music before participating in the experiment
should be rather low. Beige and blue pens are very typical to students, giving rise to a latent inhibition effect, i.e., a situation where
the CS has been encountered alone several times before it is paired
with the US. All these made it rather unlikely for classical conditioning to emerge.
The observed classical conditioning effect in Gorn (1982) might
have been due to demand characteristics, which are rather prevalent in classical conditioning studies involving human beings.
Some aspects of his procedure might have clued participants to
the study’s objectives. For example, participants were told that
the purpose was to evaluate ad music and pen color choice. The
pen color choice was obtrusively measured by telling participants
wanting different pen colors to walk to the opposite sides of the
room, hence underscoring the importance of color choice in their
minds. Furthermore, the experiment was administered in a large
class, so interaction among participants was possible and could
have affected the findings in an unpredictable manner. Finally, because no control procedure was used, alternative explanations
such as the mere exposure effect and pseudoconditioned responses
could not be ruled out.
Feinberg (1986) conducted a series of experiments in which he
manipulated the presence or absence of credit card stimuli in the
context of experimental tasks requiring participants to evaluate
the product’s worth and to indicate how much they would donate
money actually or be willing to donate to a charity. The credit card
served as an US while the product or charity activities served as
CSs. The results showed that the mere presence of a credit card
led to greater valuations and donations and that participants’ decision time was faster in the presence of a credit card symbol.
The classical conditioning effects observed in Feinberg’s (1986)
experiments were subject to some limitations. First, Feinberg did
not remove the US – the credit card itself – when measuring the
conditioned response, which means participants who viewed catalog pictures of products in the presence of credit card stimuli also
evaluated the worth of these products in the presence of a credit
card. Second, the CS lacked predictiveness of the US because a
credit card (a US) is frequently encountered alone in our daily life.
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
Third was the US pre-exposure effect – conditioning will not occur
if consumers have been preexposed to the US alone. Finally, all his
experiments employed a single simultaneous paring of the CS and
the US. While classical conditioning may occur after one trial, it
usually entails a very strong US such as an intense shock or a nauseating drug. Demand artifacts might be responsible for the observed classical conditioning effects.
The mood induction experiment in Groenland and Schoormans
(1994) is included here to demonstrate a single pairing effect with
the US preceding the CS (backward conditioning). The mood induction was operationalized in a way that only mood induction could
explain the occurrence of classical conditioning. Based on the pretest, a fragment of popular classical music (pleasant music) and a
fragment of hard rock music (unpleasant music) were selected.
Both fragments were instrumental. Participants’ personal favorite
colors were also pretested, with blue and green pens being selected
for the experiment.
Participants were split into two groups: one exposed to pleasant
music while the other to unpleasant one. The experimenters explained that a pen manufacturer wished to know whether music
could be used in a pen commercial. A 7-min musical fragment
was played to induce mood. Five seconds after the music ended,
a series of three slides of a green pen was presented for 10 s each.
Between slides was a break for 2 s. Then respondents filled a short
questionnaire, which included questions evaluating the pen shown
on the slide, and chose a pen (from a box containing many green
and blue pens in equal number) to evaluate its physical attributes.
Next, they answered another short questionnaire containing questions evaluating the music, mood, physical qualities of the pen chosen, filler questions, and others.
The results revealed that the pen evaluation was more positive
in the positive mood induction (pleasant music) than in the negative mood induction (unpleasant music), indicating that mood
influenced product evaluation. Furthermore, the expected pen color was chosen more often by participants in a positive mood induction (pleasant music) than by those in the negative mood induction
(unpleasant music), meaning mood induction influenced not only
product evaluation but also product choice.
Groenland and Schoormans (1994) used a familiar CS (a green
pen) and a familiar US (a fragment of popular classical music) with
the US preceding the CS (backward conditioning) once only. All of
these procedures are obstacles to classical conditioning.
Tom (1995) attempted to replicate Gorn’s (1982) results by
using music (positive music using Kenny G’s Song Bird song versus
negative music using John Lennon’s Number 9 Dream song) as an
US. Two-hundred and twenty-seven students saw an advertised
pen on a slide while the music was played for 60 s and had to
pay attention to the presentation (attended stimulus), during
which a neutral Chinese ideograph (unattended stimulus) was
flashed 12 times by a tachistoscope on the same screen for
0.02 s. Then, participants chose a pen and a symbol as a logo for
the pen. The results revealed that positive music led to higher
choice rates for the advertised pen and Chinese ideograph, compared to the unadvertised pen and Chinese ideograph. Nevertheless, the use of negative music as an US did not affect the
preference for either the advertised pen or advertised Chinese
ideograph. This null result will be discussed again in Section 4.
Same as Gorn’s (1982) study, Tom (1995) used a familiar CS and
US with a single simultaneous pairing of the CS and the US. The
instruction for participants to help select music for a pen ad could
have heightened the importance of both the pen and the music.
Lie et al. (2010) conducted two experiments in New Zealand to
replicate Feinberg’s (1986) study. In Experiment 1, 80 undergraduate students were randomly assigned to either task A or task B, and
within each task, they were randomly assigned to either the credit
card present condition or the credit card absent condition. Upon
285
completion of the tasks, participants answered questionnaires
measuring prior credit card use and awareness.
Task A consisted of a paper catalog of 12 consumer items, all of
which were brand-neutral and were selected because they were
similar to those used by Feinberg (1986). Each photo was presented
in the middle of a white sheet of paper, with two questions written
beneath the photo. The first question (a distracter question) asked
about the most distinctive feature of the item, while the second
question asked how much money the participant was willing to
spend on the item. Participants were individually tested and seated
at a table with a consumer products catalog placed in front of them.
For participants in the credit card present condition, a laminated
picture of the VISA and MasterCard logos was located on the upper
left corner of the table, and these participants were told the credit
card picture was left over from another experiment. For participants
in the credit card absent condition, the credit card picture was absent from the table. All participants were instructed to view each
of the catalog items and answer the two questions for each item.
Task B was a computer-based version of task A written in VisualBasic.NET and consisted of 12 trials, each of which presented one
of the consumer items (same as those in task A) in the center of the
screen. A question-and-answer box was displayed beneath each
picture, with an ‘‘Accept’’ button located beneath the answer box.
Participants were tested either individually or in groups of up to
12 persons and were presented with each of the 12 trials (i.e.,
items). For each item, participants were first asked, ‘‘What is the
most distinctive feature of the product?’’ followed by ‘‘Enter
amount you would spend ($) (only enter a single amount, not a
range).’’ Participants typed their answer and pressed the ‘‘Accept’’
button to move onto the next question or trial. The time taken between pressing the ‘‘Accept’’ button for the first question and
pressing the ‘‘Accept’’ button for the second question was recorded
as a measure of response time for the price estimation. In the credit
card present condition, a picture of four credit card logos was
shown on the bottom left-hand corner of the screen, whereas in
the credit card absent condition, the bottom left-hand corner of
the screen was blank.
The results showed that price evaluations were significantly
lower for participants who viewed the items in the presence of
the credit card symbols compared to those who viewed the items
without the symbols, and this effect emerged regardless of the
method of item presentation. These effects were taken as supporting the premise that New Zealand students’ negative conditioning
history with credit card stimuli (i.e., credit cards being associated
more with debt than spending) led to a negative credit card effect,
whereby credit cards discouraged rather than encouraged spending, and that the absence (Hunt et al., 1990; Shimp and Moody,
2000) or presence (McCall and Belmont, 1996; McCall et al.,
2004; Monger and Feinberg, 1997; Prelec and Simester, 2001; Raghubir and Srivastava, 2008) of a positive credit card effect in earlier
studies depended on past associations with credit card stimuli.
Moreover, contrary to Feinberg’s (1986) results, the mean response time for the items was not consistently affected by the
presence or absence of the credit card stimuli. Analysis of the questionnaire data on credit card use indicated that 67% of the participants reported owning or having owned a credit card, and there
was little effect of credit card use experience on price estimation
when the credit card stimulus was absent. However, a notable effect existed when the credit card was present in that participants
who did not have credit card use experience perceived the items
as having lower value. As explained by the authors, this might be
because in the absence of any personal experience where credit
cards were linked to consumption, the negative connotations associated with credit cards in the media gave rise to negative associations with credit cards. This negative association, nevertheless,
decreased with personal experience.
286
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
Sixteen year 4 students participated in Experiment 2. The procedure of Experiment 2 was identical to task B in Experiment 1 except two changes were made to the product items. Overall, the
results of Experiment 2 replicated those of Experiment 1.
Lie et al.’s (2010) results might be viewed as supporting a classical conditioning explanation of the credit card effect. In New Zealand, credit cards have acquired negative valence through repeated
pairings with negative emotions portrayed in media reports and
advertising. Therefore, credit card logos might act as second-order
CSs, with cautious spending as the conditioned response (Lie et al.,
2010). However, their observed classical conditioning effects deserve some cautions. First, same as Feinberg (1986), the study used
a single simultaneous CS–US pairing, lowering the CS’s predictiveness of the US. Second was the latent inhibition – a credit card
(serving as an US) is commonly seen in our daily life before pairing
with the product items (serving as CSs) in the experiments, resulting in little classical conditioning. In addition, participants without
credit card use experience should not be included in the study.
Alternatively, credit card use experience should be manipulated
as another independent variable. The reason for the result that
price evaluations were significantly lower for participants who
viewed the items in the presence of the credit card symbols compared to those who viewed the items without the symbols may be
that the credit card present condition happened to consist mainly
of participants without credit card use experience (because it was
found that these individuals perceived the items as having lower
value).
3.2. Studies using familiar CSs, familiar USs, and multiple CS–US
pairings
Bierley et al. (1985), in an attempt to replicate Gorn’s (1982)
findings, used colored geometric figures as CSs and music from
the movie ‘‘Star Wars’’ as an US. Student participants were divided
into four groups: (i) a ‘‘red-predictive’’ group, in which red CSs
were consistently followed by music, blue CSs were followed by
music on half of their presentations, and yellow CSs were never followed by music; (ii) a ‘‘yellow-predictive’’ group; (iii) a random
control group; and (iv) a CS-only control group. Participants were
exposed to 84 trials (28 trials per color).
Bierley et al. (1985) found that when the color of the CSs was
held constant, the CSs that were followed by music were preferred
to the ones that were not. The effect of conditioning revealed by
the group-by-color interaction was complicated. A non-significant
increase in preference for the predictive CSs (excitatory conditioned response) and a significant or non-significant decrease in
preference for the CSs combined to predict the absence of music
(inhibitory conditioned response). The group-by-color interaction
was also significant for the participants classified as unaware of
the study’s purpose or of the relation between the figures and
the music, thus ruling out the argument that awareness was necessary for classical conditioning.
Bierley et al.’s (1985) findings, nevertheless, may not be regarded as promising supports for classical conditioning. First, comparing both experimental groups with the random control group,
the increase in preference for the predictive CSs was not significant. Besides, the inhibitory conditioned response was significant
only in the yellow-predictive group. Second, like Gorn’s (1982)
study, the chance for classical conditioning to occur was likely to
be minimal due to the use of a familiar CS and a familiar US. This
US was not pretested to ensure that participants would really like
it. Although it was required that participants in the experiment like
the music from ‘‘Star Wars,’’ there was no guarantee that participants would conform to this requirement. Third, the cover story
asking participants to try to predict music might have led to
demand artifacts, which could not be detected by the post hoc
analysis if participants had poor verbalization or were not attentive
enough to describe fully what they thought.
In Experiment 1 of Janiszewski and Warlop (1993), 54 undergraduates entered a laboratory one at a time to watch soft drink
commercials and received either forward conditioning or random
conditioning. The original commercials were two 30-s TV commercials: a Mountain Dew commercial featuring white-water surfing
and a Canada Dry commercial featuring scenes of couples having
fun. For each commercial, the 18 segments arranged to catchy jingles were divided into three groups: six segments showing the
product only (serving as CSs), six segments being interesting or
fun (serving as USs), and six segments of the product being consumed (serving as fillers).
Two experimental commercials were made from each original.
The forward conditioning version of each commercial consisted
of six trials, with each trial composed of a product segment (CS),
followed by an entertaining segment (US), and then by a product-consumption segment (filler). The random conditioning version of the commercials consisted of a random order of the six
trials representing the six possible orders of the CS, the US, and
the filler segments.
The experimental design was a between-subjects manipulation
of the conditioning procedure (forward, random) with a stimulus
replication. One treatment group received 18 forward conditioning trials for Mountain Dew and 18 random conditioning trials
for Canada Dry, while the other treatment group received 18 forward conditioning trials for Canada Dry and 18 random conditioning trials for Mountain Dew. The conditioning trials for both
groups were achieved by embedding three presentations of forward and random conditioning commercials within a sequence
of six filler ads, all of which were moderately known brands to allow observation of a conditioning influence. The experimental
commercials were presented as part of a series of commercials
for eight different soft drinks. Each of the two experimental tapes
consisted of 12 commercials, each of which was separated by 5 s
of black space.
The results showed that participants receiving forward conditioning presentation of the Mountain Dew commercial looked at
the Mountain Dew container earlier than those presented with
the random version of the commercial, while participants receiving
forward conditioning presentation of the Canada Dry commercial
looked at the Canada Dry container sooner than those viewing
the random version. Neither the different preference between the
forward conditioning and the random conditioning versions of
each commercial nor contingency awareness of the CS–US pairing
(participants had little awareness of the CS–US contingency) appeared to account for the results.
The classical conditioning effects observed in Janiszewski and
Warlop (1993) were subject to some limitations. First was the US
pre-exposure effect – conditioning will not occur if people have
been preexposed to the US alone (white-water surfing scenes for
the Mountain Dew commercial and couples having fun scenes for
the Canada Dry commercial). Second was the blocking effect,
which refers to the situation when prior experience that CS1 is predictive of an US blocks conditioning from happening to CS2 and
implies that a familiar US should not be used. In addition, participants were exposed to so many commercials, leading to fatigue
and boredom. Third, using well-known brands (Mountain Dew
and Canada Dry) as CSs posed the latent inhibition effect, that is,
a situation where the CS has been seen alone several times before
it is paired with the US, resulting in little classical conditioning.
Lastly, it was not clear whether participants’ pre-existing familiarity with the experimental brand was measured. It could be that
participants in the forward conditioning happened to be more
familiar with the experimental brand, so they looked at the brand
sooner than those in the random conditioning.
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
In the affective conditioning experiment in Groenland and
Schoormans (1994), two conditions (an experimental condition
and a random control condition) were manipulated, with music
serving as an affect-producing US and a green pen serving as a
CS. Both conditions were split into two groups: one being presented pleasant music (a fragment of popular classical instrumental music pretested to be pleasant) whereas the other being
presented unpleasant music (a fragment of hard rock instrumental
music pretested to be unpleasant). In the experimental condition,
the presentation sequence consisted of seven trials. Each trial
started with a 10-s presentation of a new slide, showing the pen
in different user contexts. Four seconds after the slide presentation
began, the music was played for one minute. The time interval between the CS and the US was thus similar to that used in other classical conditioning studies. Between trials was a break for 15 s with
neither music nor slides being presented. In the random control
condition, the same procedure applied except both the slide and
the music presentation were randomly scheduled within the total
presentation period so that no sequential association between the
CS and the US could be established. The total music presentation
time was seven minutes for both the experimental and the control
conditions.
The results of the affective conditioning experiment showed a
more positive evaluation of the pen in the positive music experimental condition than in the negative music experimental condition, whereas in the control condition, no such difference was
found, indicating that affective conditioning occurred at the level
of product evaluation. For both the experimental and the control
conditions, the proportion of participants choosing the expected
color of pen (green in the positive conditions, blue in the negative
conditions) was compared to the proportion of the relevant personal favorite color (green in the positive conditions, blue in the
negative conditions). The results showed that in the experimental
condition, participants chose the expected pen color significantly
more often, while in the control condition, they did not, and this
could be interpreted that affective conditioning occurred at the level of product choice.
Groenland and Schoormans (1994) utilized a random control
condition. However, the experiment used a familiar CS (a green
pen) and a familiar US (a fragment of popular classical instrumental music). Both aspects hinder classical conditioning.
Barnes-Holmes et al. (2000) exposed 38 undergraduates to a
series of third-order conditioning experiments. In Experiment 1,
each participant was trained, tested, and did the rating task individually. In the training stage, each participant was trained on four
matching-to-sample tasks of the following types: CANCER ? VEK;
VEK ? BRAND X; HOLIDAYS ? ZID; and ZID ? BRAND Y. Participants were seated at a computer to see a text on the screen, which
instructed them to press Z if they chose the comparison on the left
of the screen and to press M if they chose the comparison on the
right. After participants pressed the space bar, one of the four
matching-to-sample tasks appeared. In effect, a sample stimulus
(e.g., CANCER) was presented for 2 s and then the screen cleared
for 0.5 s before two comparison stimuli (i.e., VEK and ZID) were
shown.
If the choice was correct, the screen cleared and the word ‘‘CORRECT’’ appeared on the screen for 1.5 s, accompanied by a highpitched tone. On the other hand, if the choice was incorrect, the
screen cleared and the word ‘‘WRONG’’ appeared on the screen
for 1.5 s, accompanied by a low-pitched tone. A 2-s inter-trial
interval was presented following the feedback, during which the
screen remained blank and no sounds were emitted. The four
matching-to-sample tasks were presented in this way in a quasirandom order until participants gave 24 correct responses
consecutively.
287
The testing stage followed the training stage immediately with
20 matching-to-sample trials, in which each of the four equivalence tasks were presented five times in a quasi-random order,
and no feedback was given after any test trial. Participants who
matched in accordance with the equivalence relations on at least
17 of these 20 trials were deemed to have passed the test. The rating stage took place right after the testing stage. A participant was
seated at a table, on which a tray contained the following: (i) on
the left side of the tray were a glass and a 1.25-l bottle of cola labeled BRAND X, and (ii) on the right side of the tray were a glass
and an identical 1.25-l bottle of cola labeled BRAND Y. Participants
tasted and then rated the pleasantness of both colas. The order of
tasting across participants was counterbalanced.
The result showed that participants who passed the equivalence
test rated the cola that had a label in an equivalence class with the
word holidays (i.e., holidays-cola) higher than the cola that had a
label in an equivalence class with the word cancer (i.e., cancercola), and no significant difference emerged between ratings of
the two colas for participants who failed the equivalence test. This
indicated that the stimulus equivalence procedure was an effective
method by which preference functions could be transferred to a
third order.
Experiment 2 tested whether the transfer of function to a third
order in Experiment 1 was facilitated by the subsequent test for
equivalence, during which the stimuli were presented contiguously on the screen. The procedure was identical to that used in
Experiment 1, excluding the equivalence testing stage. The result
showed a significant preference for the holidays-cola over the cancer-cola, suggesting the test for equivalence in Experiment 1 was
not necessary for the transfer of preference function to a third order by means of a stimulus equivalence procedure.
Experiment 3 replicated Experiment 2, but several control measures and procedures were included in order to determine whether
the conditional discrimination training clearly influenced the
pleasantness ratings of the colas. First, participants rated the two
colas before the conditional discrimination. Second, before and
after the experiment, participants rated the words ‘‘cancer’’ and
‘‘holidays’’ as emotionally negative or positive, so that these ratings
could be compared to the pleasantness ratings obtained for the colas. Third, after participants had gone through the conditional discrimination training and pleasantness ratings, they were exposed
again to the training and testing procedure except the labels Brand
X and Brand Y were swapped within the training. Overall, the results of Experiment 3 provided strong evidence that the conditional discrimination training functioned as a powerful
determinant of participants’ preference ratings.
The classical conditioning effect found in Barnes-Holmes et al.’s
(2000) called for some cautions. First, participants may have realized what the research was attempting to achieve, and the preference ratings simply reflected their tendencies to please the
experimenter. In particular, the study’s training phase was in effect
ensuring that participants were aware of the pairings. The pretest–
posttest measures, as well as the second training that swapped the
brands, in Experiment 3 would have clued participants as to what
the experimenter wanted. Second, both colas were presented
simultaneously on the tray and thus the presence of one brand
may have affected participants’ ratings of the other brand. That
is, the influence of the context stimuli could not be ignored. Finally,
the US pre-exposure effect applied – conditioning will not occur if
people have been preexposed to the US alone.
In a study designed to demonstrate that classical conditioning
effects on attitude can happen without demand characteristics or
contingency awareness, Olson and Fazio (2001) found that attitudes could develop through implicit covariation detection in a
classical conditioning paradigm. In two experiments purportedly
288
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
about video surveillance and vigilance, undergraduate females
viewed several hundred randomly presented words and images
interspersed with critical simultaneous pairings of valenced USs
[either (i) positive words such as excellent, awesome and images
such as puppies, hot fudge sundaes or (ii) negative words such as
terrible, awful and images such as a cockroach, a man wielding a
knife] with a CS (a colored picture of Pokémon cartoon character).
The findings indicated that in a covariation estimation task
involving the stimuli presented in the conditioning phase, participants revealed no explicit memory for the pairings, suggesting no
CS–US contingency awareness. In a surprise evaluation task, the
CS paired with positive items was evaluated more positively than
the one paired with negative items. This attitudinal conditioning
effect was observed when using an explicit measure in both experiments and when using Greenwald et al.’s (1998) Implicit Association Test, which was less susceptible to demand artifacts than
explicit measures, in Experiment 2.
The limitations of Olson and Fazio (2001) were twofold. First,
the study utilized a simultaneous conditioning procedure as used
in Baker (1999), Gorn (1982), Allen and Madden (1985), Rozin
et al. (1998), and Tom (1995), instead of having the CS preceding
the US as suggested by McSweeney and Bierley (1984) and Groenland and Schoormans (1994). Second, the CS (Pokémon cartoon
character) might not be novel as intended since this cartoon is
quite well-known.
Distinguishing between implicit attitudes [which are automatic
and based on associations in memory, typically measured by
Greenwald et al.’s, 1998 Implicit Association Test] and explicit attitudes (which are deliberative and based on self-reported measures), Gibson (2008) extended the studies of Shimp et al. (1991)
and Cacioppo et al. (1992), which will be discussed later. Based
on the result of a pretest pertaining to an explicit measure of preference for soft drinks, 56 undergraduate students participated in
Experiment 1, which employed three conditions: a strong preference for Coke, a strong preference for Pepsi, and an approximately
equal preference for each.
Participants were told that the experiment focused on people’s
vigilance and attention to different brand images. They were exposed to a large number of words and images on a computer screen
over five blocks, each of which consisted of 86 trials presented for
1.5 s each. Within each block, 10 of the 86 images were of the target brand, and these were randomly interspersed throughout the
block of trials. The 76 remaining trials within each block included
other neutral images and words, blank screens, and also four
images of Coke and four images of Pepsi (the conditioning trials).
Thus, over the five blocks of trials, participants saw a total of 20
images of Coke and 20 images of Pepsi, both of which were evenly
spaced across the block of trials and always alternated in order.
In the ‘‘Pepsi-positive’’ conditioning condition, Coke images
were always paired with negative photographic stimuli and words,
whereas Pepsi images were always paired with positive ones. In
contrast, in the ‘‘Coke-positive’’ condition, Coke images were always paired with positive photographic and verbal stimuli,
whereas Pepsi images were always paired with negative ones.
The paired positive or negative stimuli were presented on the same
screen as the image of Coke or Pepsi, rendering this a simultaneous
conditioning procedure. Within each level of this conditioning, the
contrasting brand was always paired with negative images and
words. In the conditioning trials, the Coke and Pepsi images were
placed on either the right or the left side of the screen, against a
black background, with the USs placed adjacent to them on the
other side of the screen. In addition, the brand images were labeled
with the word Coke or Pepsi underneath the image. Each brand appeared equally often on the right or the left of the screen.
The results of Experiment 1 showed a significant correlation
between implicit and explicit brand attitudes across the whole
sample and a significant interaction effect between the pretest attitude toward the brands and the conditioning procedure on implicit
attitudes. That is, the conditioning procedure had no effect on implicit attitudes for either the Coke-lover or Pepsi-lover groups (This
result will be presented again in Section 4). On the contrary, neutral participants showed a significant effect of conditioning such
that those in the Coke-positive condition showed more-favorable
implicit attitudes toward Coke, while those in the Pepsi-positive
condition exhibited more-favorable implicit attitudes toward
Pepsi.
As for explicit brand attitudes, participants who reported strong
pretest preferences continued with those preferences and those
who were neutral on the pretest remained neutral after the conditioning procedure. Furthermore, there was no significant main effect for the conditioning procedure and no interaction between
conditioning and pretest preference (This result will be presented
again in the Section 4). Therefore, Experiment 1 revealed that implicit attitudes (but not explicit attitudes) for mature brands could
be altered by evaluative conditioning, and this effect occurred only
for participants whose attitude toward the brand was initially neutral. In addition, contingency awareness was not necessary to
change implicit brand attitudes. The results of Experiment 2 replicated and extended the results of Experiment 1, indicating that
brand choice was congruent with conditioning but only under cognitive load, and the implicit brand attitudes predicted brand choice
but only under cognitive load.
The classical conditioning effects on implicit memory for people
with neutral attitude toward the CS found in Gibson’s (2008)
experiment had a few reservations. First was the US pre-exposure
effect – conditioning will not occur if people have been preexposed
to the US alone. Second is the blocking effect, which indicates that
familiar USs should not be used. Third is the latent inhibition – the
CS is presented alone several times before pairing with the US,
resulting in little classical conditioning.
3.3. Studies using familiar CSs, unfamiliar USs, and multiple CS–US
pairings
Stuart et al. (1987) conducted experiments with students under
ideal conditions for classical conditioning to occur. Experiment 1
used a neutral fictitious brand as a CS and pleasant pictures as
USs. The design was a 4 (levels of conditioning trials: 1, 3, 10, 20
trials) by 2 (experimental group versus random control group) ANOVA. The results showed that the experimental groups had significantly more-positive attitudes toward the brand than the
corresponding control groups. Experiment 2 utilized 10-trial and
one-trial levels with five conditions: (i) a latent inhibition group
preexposed to the CS alone several times before the CS–US pairings, (ii) a conditioning control group, (iii) a random control group,
(iv) a CS-only control group, and (v) a latent inhibition/random
control group.
The results showed that the conditioned response in the latent
inhibition groups was significantly more positive than that in the
random control groups but was less positive than that of the conditioning groups, whose responses were more positive than the
other control groups. In line with Spetch et al. (1981), Experiment
3 showed that backward conditioning did produce a conditioning
effect, relative to the control group, though the effect was significantly less than that of the forward conditioning group. Experiment 4, which examined forward and backward conditioning
procedures, confirmed that all three forward conditioning procedures resulted in similar conditioning effects, which were significantly greater than that of the control group, and replicated the
results of Experiment 3.
Stuart et al. (1987) utilized a careful design and attempted to
reduce hypothesis guessing by using filler materials. The post hoc
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
analysis, however, revealed that half of the participants were
aware of the CS–US contingency, and this contingency awareness
was a significant predictor of participants’ attitudes. This is in line
with many studies, which found that contingency awareness acted
as a causal mediator in conditioning (Allen and Janiszewski, 1989;
Baer and Fuhrer, 1982; Biferno and Dawson, 1977; Dawson, 1973;
Perruchet, 1985; Rescorla, 1988; Shimp et al., 1991). In addition,
because they did not employ an elaborate demand artifact assessment, they admitted that this possibility could not be rejected.
In another study, Walther and Grigoriadis (2004) investigated
the impact of mood on the acquisition of affective attitudes
through a 2 (mood: positive versus negative) between-subjects
by 2 (conditioning: appetitive versus aversive) within-subjects design. Forty-four students were seated individually in front of a
computer screen. The conditioning procedure contained three
sequential phases guided by a computer program: the baseline
phase, the conditioning phase, and the test phase. Participants
put on headphones and started a program of auditory instructions,
which informed them that the study was concerned with spontaneous feelings people experienced towards other objects and individuals. Participants then judged different types of stimuli
randomly selected by a computer program. In all conditions, 105
black-and-white pictures of white male faces intermixed with 67
pictures of all kinds of shoes (without brands) selected from German magazines and the Internet were displayed for 3 s each in a
full frontal view in the center of the screen. After 3 s, participants
evaluated the stimulus. Two negative USs, two positive USs, and
eight moderately rated CSs were selected from this baseline
evaluation.
Next was the mood manipulation. Participants were seated in
front of a television set. In the happy-mood condition, participants
viewed two short films: one containing scenes from the movie
‘‘Shrek’’ and the other from ‘‘The Little Bear.’’ Those in the sadmood condition were shown two films: one about child abuse
and the other about the last day of a doomed man. The films lasted
about 10 min in both conditions. After the film presentation, participants rated their mood on a 9-point scale embedded in several
neutral questions concerning the film.
The conditioning phase followed the mood manipulation. Pictures of neutral shoes (CSs) were paired with liked or disliked
faces (USs). Two sets of stimuli were used, each of which consisted of a CS-positive US pair, a CS-negative US pair, and a control CS–CS pair. Identical stimuli were presented in each pair.
Both sets were presented five times, intermixed with each other
and with two additional neutral CS–CS distracter pairs. In keeping
with previous studies, each picture was shown for 1 s with a trace
interval (i.e., the interval between the end of the first stimulus
and the beginning of the second stimulus of a pair) of 1 s and
an inter-trial interval of 4 s. The conditioning phase was instantly
followed by the test phase, which employed the same procedure
as in the baseline. A second mood check was conducted immediately after the test phase, and then the recognition test appeared
on the screen.
Participants saw each CS in its original size on the left side of
the screen and with four smaller pictures on the right. The instructions above the pictures asked participants to select the small picture that frequently followed the large picture. Participants had the
choice between the correct US, a stimulus with the same valence as
the correct US, a stimulus that occurred with the same frequency
as the correct US, and a stimulus of medium valence selected from
the baseline. Once they made their choice, the next CS recognition
test appeared on the screen. Finally, participants answered an
open-ended questionnaire to check their demand awareness and
verbal knowledge of the contingencies. After completing the questionnaire, participants filled out a final mood check. The procedure
lasted for about 45 min.
289
The results revealed an interaction between the valence of the
US (liked versus disliked faces) and participants’ mood. That is, participants in a sad mood were more susceptible to the negative conditioning procedure than were those in a happy mood. In addition,
contingency awareness influenced the formation of consumer
attitudes.
Walther and Grigoriadis (2004) used a very tedious conditioning procedure and an extensive set of open-ended questions to
check contingency and demand awareness. After about 40 min of
going through the experiment and several rounds of answering
many questions, it was doubtful how participants could answer
the questions accurately, and they might have been too exhausted
to elaborate on the awareness-checking questionnaire.
3.4. Studies using unfamiliar CSs, familiar USs, and multiple CS–US
pairings
Additional evidence in favor of classical conditioning was from
Allen and Janiszewski (1989). Experiment 1 assessed the role of
contingency awareness versus demand awareness by using five
neutral Norwegian words as CSs and positive evaluative phrases
for participants’ performance as USs. The five words had different
degrees in predicting the US. The experimental group was found
to have significantly more-favorable evaluation for the words predicting the US in both the between-group test and the withingroup test.
Experiment 2 attempted to strengthen the conditioning procedure by increasing inter-trial intervals, to heighten contingency
but not demand awareness, and to manipulate demand awareness.
The results showed that in the extended inter-trial interval group,
there were no effects on word evaluations or brand name ratings.
In the contingency-aware group, there was an effect on word evaluations only. In the demand-aware condition, the effects on both
variables were significant. Taken together, the two experiments
did not support the conditioning-without-awareness position
(Kassarjian, 1986; MacKenzie et al., 1986; Preston, 1982; Staats
and Staats, 1959).
In Shimp et al. (1991), the CS in each of the 21 forwardconditioning experiments was one of the four unknown
(Cragmont, Elf, My-te-Fine, and Target), two moderately known
(RC and Shasta), or two well-known (Coke and Pepsi) cola brands.
The US was a composite of four attractive water scenes used in Stuart et al.’s (1987) study (a mountain waterfall, a sunset over water,
a boat mast against the sky, and a lavender-hued island). The filler
context for the conditioning trials included three unknown or
known brands. Specifically, when an unknown brand served as
the CS in the context of other unknown colas, the filler items comprised the three remaining unknown brands. When an unknown
brand was conditioned in a known context, Coke, Pepsi, and either
RC or Shasta were used as fillers. When a moderately known brand
was conditioned in the context of known brands, the filler brands
included Coke, Pepsi, and the remaining moderately known brand.
When a well-known brand was conditioned in the context of
known brands, the filler brands were composed of RC, Shasta,
and the remaining well-known brand. When a moderately or
well-known brand was conditioned in the context of unknown
brands, any three of the four unknown brands served as fillers.
Each experiment involved a conditioning group and a random
control group. The conditioning group received 20 conditioning
trials, in which the CS always preceded an US scene, and 60 nonconditioning trials, in which non-CS brands of cola were paired
with 12 pictures pretested as being neutral, e.g., a license plate,
weeds growing in a pond, and unpainted boards. All conditioning
and non-conditioning trials involved displaying a cola brand on a
screen for 7.5 s followed by a 7.5-s slide of either a positively valenced slide in the case of the conditioning trials or a neutral slide in
290
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
the case of non-conditioning trials. All 80 15-s pairings ended with
a 2-s dark-screen pause. Each of the four US water scenes followed
the CS five times, with the various water scenes randomly distributed among the 20 conditioning trials.
To minimize temporal conditioning, the 20 conditioning trials
were interspersed among the 60 non-conditioning trials. The time
between the end of a conditioning trial and the onset of the next
trial (i.e., the inter-trial interval) ranged from 2 to 102 s with an
average of 54 s. The random control group received the same number of presentations of the CS and the US and the same number of
presentations of the filler brands and scenes, but all were randomly
assigned with respect to each other. The 23-min slide presentation
was interrupted twice after the first and the second portion of the
three approximately equal portions for participants to answer
questions about attitudes toward filler brands.
The 21 experiments ranged in size from 51–83 students, with
most experiments having approximately 30 participants each in
the conditioning and random control groups. The experimental
sessions consisted of 2–10 students, with most sessions having
4–7.
The results showed that the conditioning groups in 11 of the 21
experiments exhibited significantly more positive attitudes toward
the CS brands than the random control groups (The non-significant
results will be discussed again in Section 4). Of the 11 instances of
statistical significance, seven employed unknown brands, three
used moderately known brands, and one used well-known brands
as CSs. Attitude change was most apparent when participants noticed the contingency between the CS and the US. In summary,
evaluative conditioning could change attitudes for novel brands
but not for mature brands (like Coke and Pepsi).
One-hundred and sixty-nine participants were recruited under
the guise of evaluating audiocassette tapes of business programs
in Blair and Shimp’s (1992) experiment, which involved two
phases: a preconditioning phase followed by a conditioning phase.
In the preconditioning phase, participants returned to the research
site many times in winter (unpleasant experience) and each time
listened to different audiotaped business books introduced and
concluded with the same music to be used as the US later. After
the preconditioning phase, about half of the participants formed
two conditioning groups, namely, the non-US-preexposureconditioning group (the conditioning group that was not preexposed to the US) and the US-preexposure-conditioning group
(the conditioning group that was preexposed to the US), while
the remainder constituted two control groups. The CS was a
neutral fictitious brand name of sportswear ‘‘Garra’’ and the US
was the ‘‘Thanksgiving’’ music positively evaluated and relatively
unfamiliar to the participants.
The entire presentation in the conditioning procedure included
a shirt slide for 4 s, a word slide for 4 s (including the CS and three
filler Portuguese names), and then a blank screen while music was
played for 25 s. This sequence was repeated 40 times. The 25-min
presentation was designed such that 10 conditioning trials were
embedded among the pairings of filler names and music snippets.
In the conditioning groups, each word slide was followed by a 25-s
snippet of the US music, and the US always followed the CS but
never any of the other names. Participants in the control groups received an identical number of exposures to the CS, US, and filler
materials as those in the conditioning groups; however, the CS
was never immediately followed by the US in the control groups
but instead was always followed by one of the three filler snippets.
The results showed no significant difference between the nonUS-preexposed conditioning group and its control. In other words,
the US was not salient enough to produce positive attitudinal conditioning for participants in the non-preexposed group (this null
result will be presented again in Section 4). Participants in the USpreexposed conditioning group evaluated the CS less positively than
those in the control group did. Besides, the non-US-preexposed
conditioning group held significantly more-favorable attitudes toward the CS than the US-preexposed conditioning group. In summary, the results indicated that the US induced negative
attitudinal conditioning toward the name Garra in participants
who had been preexposed to the US in an unpleasant context.
Blair and Shimp’s (1992) study used proper control procedure
but faced some limitations. In the US-preexposed group, the supposedly pleasant thanksgiving music in the preconditioning phase
was in effect a CS being paired with unpleasant winter experience
until this music was later used as a negative US. However, this preconditioning procedure gave rise to the blocking effect – the US
should have no previous associations with the CS or certain experience before. It would have been better if the negative US was
manipulated via the use of music pretested to be unfamiliar and
unpleasant to people similar to the target respondents. Second
was the Garcia effect – not any stimulus can be used as a CS or
an US, and the CS and the US should logically belong together in
a classical conditioning experiment. Thanksgiving music (the US)
and sportswear (the CS) might not logically/emotionally belong together. The negative classical conditioning effect found might have
been caused by some other extraneous factors.
In Experiment 1 of Kim et al. (1996), 66 undergraduate students
were administered in groups of 2–5 persons and randomly assigned to one of the three conditions, namely, forward conditioning, backward conditioning, and control. A picture of a plain
white pizza box inscribed with the logo of a fictitious brand ‘‘L Pizza House’’ served as a CS, while a race car, which was pretested to
provoke neutral affective response and convey implicit meaning
that might be used in forming beliefs about the brand, served as
an US. Meanwhile, 60 pictures in the series of 80 images for three
other fictitious brands were used to detract attention from the focal CS–US pairings and thus decrease hypothesis guessing.
Participants in each condition watched a series of 80 visual
images displayed on a television. In the forward conditioning
group, they were exposed to ten trials, each of which consisted
of a 7.5-s presentation of the CS followed by a 7.5-s presentation
of the US. In the backward conditioning group, all the ten presentations of the CS were preceded by the US. Participants in the control group were exposed to the same number of presentations of
the CS and the US and the same 60 filler images as those in the
treatment groups. However, the sequence of images was randomly
scrambled, and the CS and the US never appeared contiguously.
The results showed that the forward conditioning group had
more-favorable attitudes toward the brand than the control and
the backward conditioning groups, which did not differ from each
other. Participants who experienced forward (versus backward)
conditioning trials demonstrated higher levels of contingency
awareness, and those who were aware (versus unaware) of the
contingency between the CS and the US exhibited more-positive
beliefs about the target attribute and more-favorable brand attitudes. Furthermore, the series of regression analyses showed that
the conditioning procedure accounted for significant variation in
beliefs about the target attribute and brand attitudes. Lastly, beliefs
about the target attribute mediated the effect of the conditioning
procedure on participants’ brand attitudes.
Experiment 2, which used a different CS (Brand L facial tissue)
and USs (pictures of a single kitten and two kittens, respectively)
from Experiment 1, confirmed the results in Experiment 1. Together, the results revealed that brand attitudes could be conditioned using both attractive images that induced direct affect
transfer and descriptive visual images that elicited inferential belief formation (i.e., cognitive mechanisms).
Kim et al. (1996) used filler brands to reduce hypothesis guessing and a control group for comparison. However, the obtained classical conditioning effects might be subject to some reservations.
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
The USs in both experiments (a race car in Experiment 1 and kittens
in Experiment 2) are typically encountered alone in people’s daily
life or television. According to the US pre-exposure effect, conditioning will not occur if people have been preexposed to the US
alone. Demand artifacts, despite the use of filler brands, might have
given rise to the observed classical conditioning effects.
Kim et al. (1998) exposed 36 undergraduates to a classical conditioning experiment, in which the fictitious restaurant ‘‘L Pizza
House’’ served as a CS and a kitten picture, which was chosen after
extensive pretesting that the picture was neutral and bore no belief
about the CS, served as an US. Participants were randomly assigned
to one of two conditions (conditioning or control). In the conditioning group, participants viewed the CS for 5 s, followed by a 5-s presentation of the US. After 10 repetitions of the CS–US pairing, 1 s of
‘‘down time’’ occurred, followed by a filler CS and a neutrally affective US in order to detract participants’ attention from the focal CS–
US pairing. In the control condition, participants were exposed to
the same stimuli as in the conditioning group; however, the order
of stimuli presentation was randomized and there was no systematic pairing of the CS and the US.
The result of Experiment 1 showed that attitude toward the
product in the conditioning group was significantly morefavorable than that in the control group, indicating that the
conditioning procedure influenced attitude toward the product.
The differences in the mean values between the conditioning and
the control groups for all product beliefs were not significant,
showing that the US (i.e., a kitten picture) did not provoke any
beliefs that could be associated with the CS (i.e., L Pizza House).
In addition, the mean values for affect toward the kitten in the conditioning and the control groups were not significantly different,
suggesting that participants’ positive affect toward the kitten was
present in both conditions.
The design of Kim et al.’s (1998) study was rather simple with
two groups only. The limitation in their study was similar to Kim
et al.’s (1996), namely, the US pre-exposure effect, which suggests
that conditioning will not occur if consumers have been preexposed to the US alone. A kitten is frequently encountered in daily
lives. Second, the Garcia effect stipulates that the US and the CS
should be matched properly. It was not clear whether a kitten is
an appropriate match as an US for a pizza house.
In Experiment 1 of Grossman and Till (1998), groups of 12–25
participants were exposed to a fictitious brand of mouthwash (Garra), which served as a CS, and three highly favorable scenes (a tropical scene with a boat, a picture of railroad tracks leading to a
snow-covered mountain, and a nature scene with a panda), which
were selected as USs based on a pretest. Nine neutral pictures were
paired with each of the three filler brands, which were included to
disguise the nature of the study and prevent hypothesis guessing.
The treatment and control groups were exposed the same number of times to both the US and the CS; however, only the treatment group was exposed to the CS–US contingency pairing. That
is, the control group saw a random sequence of pictures, whereas
the treatment group saw each of the following stimuli for 4 s in
the following order: the CS, the US, the CS superimposed on the
US, and a blank screen. After the experiment, participants completed an immediate attitude measure, a one-week delayed measure with the same questionnaire as the immediate one, and a
three-week delayed measure with identical questions as the former ones but in different order, respectively.
The results showed that the treatment participants’ mean attitude toward the CS was much higher than that of the control participants. Moreover, the significant difference of mean attitude
toward the CS between the treatment and the control conditions
persisted over time as reflected in the one-week and the threeweek delayed conditions, indicating that participants exposed to
a classical conditioning procedure in which a CS was systematically
291
paired with a positive US maintained a favorable attitude toward
the CS over time. The result in Experiment 2 also showed that conditioned attitudes toward the brand did persist over time. The classical conditioning effects obtained in Grossman and Till (1998)
were subject to the same reservations as in Kim et al. (1996).
Priluck and Till (2004) conducted a 2 (conditioning group versus
control group) by 2 (high versus low involvement) betweensubjects factorial-design experiment with 195 undergraduate students. Similar to Grossman and Till (1998), the CS was a picture of
a fictitious brand of mouthwash (Garra) that had been previously
pretested to be affectively neutral and not to evoke unusual associations. The USs were three pleasant visual scenes: a picture of a
boat in tropical waters, a picture of railroad tracks leading to a
snow-covered mountain, and a picture of a nature scene with
a panda. Each of these pictures had been pretested to be favorable
to the participants.
Participants were exposed to a positive conditioning procedure
in which the test brand was paired with favorable stimuli either 1
or 15 times. Participants in both the conditioning treatment and
the conditioning control groups were exposed to a slide presentation in which the stimuli appeared interspersed among filler pictures. While the conditioning control group saw a random
sequence of pictures, the conditioning treatment group was exposed to each of the following stimuli for 4 s in the following order:
the CS, the US, the CS superimposed on the US, and a blank screen.
High involvement was achieved by written instructions asking participants to pay careful attention to the slide presentation with a
chance to win $25 if they answered questions about the presentation correctly. In contrast, low involvement was attained by asking
participants to consider filler pictures.
The results of Experiment 1 showed that the mean attitude toward the CS among contingency-aware participants in the conditioning treatment was significantly higher than that among
contingency-unaware participants. The three-step regression procedure supported the hypothesis that contingency awareness mediated the relationship between classical conditioning procedures
and attitudes toward the CS because: (i) participants in the conditioning treatment were more likely to be aware of the CS–US contingency than those in the conditioning control; (ii) attitude
toward the brand was significant when regressed on conditioning,
showing a basic conditioning effect; and (iii) when attitude toward
the brand was regressed on awareness and conditioning, the effect
of conditioning decreased. Finally, individuals exposed to the CS–
US pairing who were both highly involved and high in need for
cognition were more aware of the CS–US contingency relationship
than were those who were not highly involved and not high in
need for cognition.
Experiment 2 revealed further that individuals exposed to the
CS–US pairing who were both highly involved and high in need
for cognition developed more-favorable attitudes toward the CS
than those who were not highly involved and not high in need
for cognition. Participants exposed to the CS–US pairing who were
highly involved or high in need for cognition developed stronger
beliefs than those who were not highly involved or low in need
for cognition. In addition, contingency awareness mediated more
strongly for beliefs than for affect transfer in the relationship between classical conditioning and attitude toward the CS. The findings in Experiment 2 were consistent with those of Kim et al.
(1996, 1998), challenging the traditional view of classical conditioning as automatic, non-cognitive learning. However, the classical conditioning effect observed in Priluck and Till’s experiments
are subject to the same reservations as in Grossman and Till
(1998) and Kim et al. (1996).
Stahl et al.’s (2009) results supported the critical role of contingency awareness in evaluative conditioning, albeit valence awareness, not identity awareness of the US. Sixteen university students
292
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
participated in Experiment 1, in which two sets (set A and set B) of
five neutrally evaluated, pronounceable nonwords were used as
CSs, and two sets of 25 pictures from the International Affective
Pictures System were used as USs. The design was 2 (US valence:
positive versus negative) by 2 (CS set: set A and set B, both are positive) with repeated measures on the US valence factor. During the
experiment, five USs were randomly assigned to each CS for each
participant anew, creating 50 different CS–US pairs.
The experiment was conducted in individual computer-controlled sessions in three phases: conditioning, evaluative ratings,
and awareness check. In the conditioning phase, participants
watched 100 CS–US pairings (50 different CS–US pairs, each presented twice). The CS and US simultaneously appeared on the computer screen for 2,000 ms with the US picture in the upper half of
the screen and the CS in the lower half. The presentation order was
randomized anew for each participant.
After the conditioning phase, participants evaluated each nonword CS on a scale ranging from 1 (very unpleasant) to 8 (very
pleasant). Then, valence and identity awareness was assessed for
each CS–US pair. Participants indicated for each CS whether they
thought it had been paired with pleasant or unpleasant USs. Meanwhile, awareness for the US identity was probed through recognition tests five times for each CS, once for each of the five USs with
which it was paired.
The results showed that for 15 out of 16 participants, valenceaware CSs were obtained; for these valence-aware CSs, a significant evaluative conditioning effect was observed across participants. For 11 out of 16 participants, valence-unaware CSs were
obtained, and evaluative conditioning was absent when only valence-unaware CSs were considered. Furthermore, a regression
analysis revealed that identity awareness did not predict evaluative conditioning for both valence-aware and valence-unaware
CSs. In contrast, valence awareness significantly predicted evaluative conditioning.
Twenty-eight university students participated in Experiment 2,
in which the same USs as in Experiment 1 were used. However, ten
product pictures completed the set of CSs, five of which were
paired with positive USs and the remainder of which were paired
with negative ones. The procedures of Experiment 2 were identical
to Experiment 1 with two exceptions. First, four evaluative ratings
(global impression, attractiveness, pleasantness, and willingness to
buy) were collected instead of only one. Second, the valence-memory test and the identity-memory test both included an additional
‘‘don’t know’’ response option.
The results showed that in the US identity test, the correct response was selected in 17.3% of cases. In the valence awareness
test, the correct response was selected in 57.1% of cases. For 26
participants, valence-aware CSs were present, and for these valence-aware CSs, a significant evaluative conditioning effect was
observed. For 13 participants, valence-unaware CSs were present,
and for these valence-unaware CSs, evaluative conditioning was
absent. The same applied to valence-undecided CSs. Identity
awareness did not predict evaluative ratings for valence-aware, valence-unaware, and valence-undecided CSs.
Experiments 3 and 4 obtained similar results as those in the
first two experiments. All of the experiments confirmed the important role of awareness (defined as recognition memory for the CS–
US pairings) for the emergence of evaluative conditioning effects
and supported the view that the impact of valence awareness on
evaluative ratings observed across all the experiments reflected
genuine valence acquisition rather than demand effects or other
artifacts of the explicit evaluation process.
Because the evaluative conditioning found in Stahl et al.’s
(2009) research entails awareness of the CS–US contingencies, it
may be viewed as classical conditioning. Some cautions apply to
the obtained effects. First, the experiments required tremendous
effort from the participants, especially in the valence and identity
awareness tests. The answers given might be just random guesses.
Second was the US pre-exposure effect – conditioning will not occur if people have been preexposed to the US alone. Third is the
blocking effect, which indicates that familiar USs should not be
used. Lastly, the study presented CS and US simultaneously on
the computer screen instead of having the CS preceding the US.
3.5. Studies using unfamiliar CSs, unfamiliar USs, and multiple CS–US
pairings
Cacioppo et al. (1992) conducted Experiment 1 on 43 students
using a 2 (male versus female) by 3 (CS–US contingency: word followed by electric shock, nonword followed by electric shock, word
and nonword paired randomly with electric shock) by 2 (experimental stimulus: word, nonword) mixed-model factorial design
in which biological sex and CS–US contingency served as between-subjects factors and experimental stimulus served as a
within-subjects factor. After completing a pretest, each participant
was tested individually. Each participant determined the annoying
but not painful intensity of the electric shock to be used as the US.
The experimental session consisted of eight presentations of a
single word and eight presentations of a nonword. Each experimental stimulus presentation lasted 7 s, and the inter-stimulus
interval was randomly varied between 30 and 40 s. In the wordshock condition, the offset of each word presentation was followed
by a 0.25-s electric shock to the calf (US), whereas the nonword
was never followed by electric shock. Likewise, in the nonwordshock condition, the offset of each nonword presentation was followed by a 0.25-s electric shock to the calf (US), whereas the word
was never followed by electric shock. In both conditions, the order
of the 16 experimental stimulus presentations was randomized.
In the random shock (pseudoconditioning) condition, the 16 7-s
experimental stimulus (word and nonword) presentations were
randomly ordered with a 30–40 s inter-stimulus interval. A 0.25s electric shock followed the offset of either the word or nonword
on eight of the 16 experimental trials. On which eight trials the
shock was executed was randomly determined. After the last stimulus presentation, participants rated the pleasantness and familiarity of the same 15 six-letter words and 15 six-letter nonwords in
the pretest.
In line with Shimp et al. (1991), Cacioppo et al. (1992) found
that evaluative conditioning was most effective for unfamiliar
CSs (unfamiliar words). In Experiment 2, 22 students read descriptions of the experimental stimuli and the CS–US contingency used
in Experiment 1 in a 2 (experimental stimulus: word, nonword) by
3 (CS–US contingency: word followed by shock, nonword followed
by shock, random pairings of word and nonword with electric
shock) by 2 (experimental instruction: predict participant’s rating,
predict experimenter’s expectation) within-subjects factorial design. The results of Experiment 2 revealed that although contingencies between the CSs and the USs were explained and participants
tried to predict the experimenter’s hypothesis, they were unable to
predict accurately the differential attitude change observed in
Experiment 1.
Schemer et al. (2008) exposed 78 undergraduates to an evaluative conditioning experiment, in which previously neutral and unknown brands represented CSs, and the actors who were not
known to the broad public in a music video functioned as a positively or negatively valenced US. The 2 by 2 between-subjects factorial design contained conditioning valence obtained through the
affective image of rap actors in a music video (positive versus negative image) and occurrence of brand placements (placement present versus absent, and participants were randomly assigned to one
of the four conditions.
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
Understanding that the study dealt with viewers’ evaluation of
rap music videos, participants read a newspaper article, which
manipulated the image of the rap actors. Then they viewed the
rap video on a personal computer for slightly more than four minutes. Equipped with headsets, they were administered in groups of
6–8 individuals in an advertising laboratory with no disturbance
among participants. After watching the clip, they filled an online
questionnaire consisting of measures of manipulation checks,
dependent variables, music preferences, and demographics. In
the placement version of the clip, the placement of the target brand
occurred 29 times simultaneously with the rap actors. The single
appearance duration did not exceed 2 s.
The manipulation checks showed that participants reading
about negative characteristics of the artists evaluated the artists
more negatively than did those reading about positive characteristics, indicating a successful manipulation of the image of the rap
actors (or the conditioning valence). The results showed that negatively conditioned individuals had a more negative attitude toward the brand in the placement condition compared to those
who saw the rap video without placement, while positively conditioned participants had a more positive attitude toward the brand
in the placement condition compared to those who saw the rap video without placement. In contrast, in the no-placement condition,
the attitude toward the brand did not differ as a function of conditioning valence, indicating that the attitudes toward the brand improved as a function of the pairing of a brand with favorable rap
actors, and the evaluation of a brand decreased when the brand
co-occurred with negatively evaluated rap actors. Moreover, attitudinal conditioning effects were stronger in individuals with higher
preference for rap music, suggesting that participants liking rap
music could be conditioned both positively and negatively. Finally,
evaluative conditioning effects were more likely to occur when
viewers did not recognize the brand embeds in the rap video.
Some reservations applied to the evaluative conditioning effect
found in Schemer et al. (2008). First, the Garcia effect suggests that
the US and the CS should be matched properly. Not any stimulus
can be used as a CS or an US, and they should belong together logically. Second, the study utilized a simultaneous conditioning procedure instead of having the CS preceding the US.
4. Evidence against classical conditioning effects
Due to the small number of studies in this section, they are arranged by publication year instead of the methodological classification scheme used in Section 3. Several studies did not find
support for the classical conditioning effects. Allen and Madden
(1985) adapted Gorn’s (1982) procedures by processing participants individually, using a different pen choice method, conducting a more systematic post-experimental inquiry, using humor as
an US, using a green and a black pen as CSs, and adding a buy-back
measure. They could not replicate Gorn’s results. Only in the liked
US condition was there an effect significantly different from a random choice, but Allen and Madden (1985, p. 309) attributed it to
the demand awareness among participants. There were no relationships between the color-selection and buy-back variables.
The results of Allen and Madden were not surprising because
the design was not much stronger than Gorn’s (1982). The experimental conditions were difficult for classical conditioning to occur
due to a single simultaneous pairing of the CS and the US, the use
of familiar CSs, and possibly a Garcia effect (i.e., an improper match
between the CS and the US). Perhaps the most crucial factor for the
lack of any conditioning effect was the fact that participants revealed very little contingency awareness of the CS and the US. This
furnished another support for the above-mentioned view that classical conditioning is mediated by contingency awareness.
293
In testing the effect of attitude toward the ad on attitude toward
the brand from a classical conditioning perspective, Gresham and
Shimp (1985) paired an advertised brand with an affectively valenced TV commercial. It was found that only a small portion of the
positive- and negative-affect ads showed attitude toward the ad
to be a predictor of attitude toward the brand, thus failing to support the classical conditioning mechanism. Negative-affect ads
seemed to elicit more of a classical conditioning response than
did positive-affect ones. The hypotheses, which were designed to
rule out mechanisms other than classical conditioning, were not
supported. Part of the inability to support the classical conditioning mechanism in Gresham and Shimp’s study might be traced to
the use of familiar TV commercials for mostly mature brands in
established product categories. Thus, attitude toward the brand
might influence attitude toward the ad, not the other way around.
Macklin (1986) exposed preschool children thrice to either an
orange or a yellow pencil under three conditions (forward conditioning, simultaneous conditioning, and random control) with a
picture of Smurf as an US. No conditioning effect was found –
the color selected did not relate to the buy-back variable, replicating Allen and Madden (1985). The lack of support for any classical
conditioning effect in Macklin’s study might be due to insufficient
number of pairings, a Garcia effect, an US pre-exposure effect, a
blocking effect, a latent inhibition effect, a small sample size, and
absence of contingency and demand awareness in children
participants.
Another case against the classical conditioning theory was by
Kellaris and Cox (1989). Experiment 1 followed Gorn’s (1982) procedures but used classical music, a yellow pen, a less obtrusive
choice procedure, and a new cover story. It used a 2 (liked versus
disliked music) by 2 (Gorn’s versus new cover story) factorial design. The music appeal was found to have no effect on choice
behavior. The cover story treatment did not affect the guessing of
true hypotheses. Experiment 2 examined the role of demand artifacts in Gorn’s experiment. Treatments like those used by Gorn
were described rather than administered, as suggested by Sawyer
(1975). Strikingly, Gorn’s results were replicated without actual
treatments. Twenty-nine percent of the participants asked to
report the purpose of the study correctly guessed the main
hypothesis.
Experiment 3 followed Gorn’s (1982) study with an equally
strong music appeal, using a 2 (high versus low music appeal) by
2 (Gorn’s cover story and choice procedures versus more disguised
procedures) design. Neither procedures yielded significant relationships between the music appeal and the pen choice. However,
there was a significant relationship between the procedure and the
guessing of true hypotheses in that 12 of the 15 participants who
correctly guessed the hypothesis had received Gorn’s procedure.
Kellaris and Cox (1989) was one of the apparent demonstrations
of demand artifacts in classical conditioning literature although
some drawbacks existed in the experiments. The critical choice
measurement would be void if participants did not follow the
experimenter’s suggestion of trying the new pen in answering
the questionnaires (the ink color differed to designate which pen
color the participants had chosen) or if the color participants intended to pick up had all gone from the box. Furthermore, because
the procedures were similar to Gorn’s, there was little chance that
classical conditioning could occur.
As presented earlier, the results of Shimp et al. (1991) showed
that the conditioning groups in 11 of the 21 experiments exhibited
significantly more positive attitudes toward the CS brands than the
random control groups while the remaining 10 experiments did
not obtain significant classical conditioning effects. Blair and
Shimp (1992) found no significant difference between the nonpreexposed conditioning group and the control group. In
other words, the US was not salient enough to produce positive
294
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
attitudinal conditioning for participants in the non-preexposed
group. Tom (1995) found that the use of negative music as an US
did not affect the preference for either the advertised pen or advertised Chinese ideograph. One of the explanations for this null effect
of the negative US [which was in line with the findings of Allen and
Madden (1985)] offered by Tom was that the negative music employed was not only disliked but also perceived to be incongruent
for the stated purpose.
Shimp and Moody (2000) exposed 69 undergraduates to three
conditions, namely (i) Feinberg (1986) condition – VISA card stimulus present both at product encoding and rating, (ii) VISA card absent – VISA card stimulus present during encoding of product
information but not at rating, and (iii) control – VISA card stimulus
absent during both encoding of product information and rating.
The VISA card served as an US. Participants were told they were
participating in a study simulating a new catalog order center
method of shopping. No significant effect was found. The absence
of classical conditioning effect in Shimp and Moody’s study might
be due to the US pre-exposure effect and the CS lacking predictiveness of the US. Besides, participants revealed very little contingency and demand awareness.
Finally, as presented earlier, Gibson (2008) found that the
conditioning procedure had no effect on implicit attitudes for
either the Coke-lover or Pepsi-lover groups. As for explicit brand
attitudes, participants who reported strong pretest preferences
continued with those preferences and those who were neutral on
the pretest remained neutral after the conditioning procedure.
Furthermore, there was no significant main effect for the conditioning procedure and no interaction between conditioning and
pretest preference.
5. Conclusion
This study reviews empirical research regarding classical conditioning effects on consumer behavior. The majority of the reviewed
studies reported significant results supporting classical conditioning effects. This is not surprising given the fact that studies with
significant findings have a higher chance to appear in international
refereed journals than studies with null effects.
The review shows that in cases where classical conditioning effects were found, the results were rather dubious due to deficiencies in methodology and/or possible demand artifacts. In most of
these studies, CS–US contingency awareness was necessary for classical conditioning effects to emerge. Alternatively, the effects could be
explained by some other mechanisms such as the mere exposure
effect and pseudoconditioned responses if no proper control procedures were employed. In cases where classical conditioning effects
were not supported, the non-significant results could be attributed
to violation of the conditions for classical conditioning to occur, absence of CS–US contingency awareness and demand awareness, or
lack of statistical power due to a small sample size.
Based on the above evaluation, it seems legitimate to conclude
that to date, not much convincing evidence exists for classical conditioning effects in consumer behavior and advertising contexts.
Whether this is because (i) the theory itself is false (although a theory cannot be falsified by lack of evidence), (ii) the theory is valid
but cannot be extended to affective responses in consumers who
are far different from animals on which the theory is based, (iii)
the methodology for testing and measuring classical conditioning
effects on consumer behavior has not been up to the task, or (iv)
classical conditioning effects on humans may occur under specific
boundary conditions only, remains to be uncovered. As raised by
Janiszewski and Warlop (1993), one reason underlying the inconsistencies in consumer conditioning literature is the flexibility in
human learning mechanisms. Humans are very sensitive to the
procedures employed to communicate information and the measures of the effect of these communications. Small procedural
changes can significantly affect the learning that is obtained in a
session. The findings of each study thus depend on the selection
of the CS and the US, the strength of the unconditioned response,
inter-trial interval, temporal priority of the CS and the US, number
of trials, training environment, test environment, test distracters,
CS consistency, control conditions, subject populations, and so
forth.
Even though classical conditioning effects might really exist,
they might not provide much usefulness in marketing given that
the affective response changes in consumers do not necessarily
lead to actual purchases and the ideal conditions for classical conditioning effects to occur are rather difficult to meet in real life.
Specifically, the advertised brand (the CS) will have low predictiveness of the US and thus low chance to elicit the desired unconditioned response if (i) both the CS and the US are presented
simultaneously (which is the case in print ads and even in some
television commercials); (ii) the CS is presented constantly, followed occasionally by the US (consumers are likely to encounter
the brand many times in real life without the presence of the US
portrayed in the ad); and (iii) either the US or the CS is frequently
encountered alone (which is usually the case in consumers’ daily
life). Marketers who are keen to utilize a classical conditioning procedure to elicit certain consumer responses should therefore be
aware of its limited chance of success in real marketplace.
Most studies reviewed shared the following weaknesses: (i)
using familiar CSs and USs, thus constituting the CS lacking predictiveness of the US, the blocking effect, the US pre-exposure effect,
and the latent inhibition effect; and (ii) lacking any pretest to ensure that the CS and the US match each other logically and perceptually. Some studies (e.g., Feinberg, 1986; Gorn, 1982; the mood
induction experiment of Groenland and Schoormans, 1994; Lie
et al., 2010) employed a single pairing of CSs and USs and yet found
significant classical conditioning effects. While classical conditioning might occur after one trial, it usually entails a very strong US
such as an intense shock or a nauseating drug, which has hardly
been used in classical conditioning research in consumer behavior
for ethical reasons.
Given the prevalent methodological deficiencies and strong
demand artifacts of past studies, future research testing classical
conditioning effects on consumer behavior should try to improve
all of the following aspects within the same experiment to increase
internal validity of the research:
(i) Making sure that the CS predicts the US by using unfamiliar
CSs and USs and by having the CS precede the US.
(ii) Pretesting that the CS and the US match each other logically
and perceptually.
(iii) Using striking CSs and USs and allowing longer inter-trial
intervals.
(iv) Including various filler materials to reduce participants’
hypothesis guessing.
(v) Including a random control group, which is exposed to the
same number of CSs and USs as the experimental group
but these stimuli are presented randomly with respect to
each other.
(vi) Including a demand-artifact-checking group, which does not
receive the same treatment as the experimental group but
merely reads the description of the treatment, in order to
compare the results with the experimental group.
(vii) Measuring CS–US contingency awareness and hypothesis
guessing in every group.
(viii) Using both negative and positive USs to see whether classical conditioning effects are present in both.
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
(ix) Manipulating the number of CS–US pairings and the intertrial intervals to see at what levels classical conditioning
effects can be observed.
Once studies with more valid methodology have been
conducted and enough supporting evidence has been accumulated,
future research can then proceed to examine the boundary conditions for classical conditioning effects to emerge and not to
emerge. Cross-cultural experiments may also be conducted to
investigate whether cultural upbringing prompts people to respond differentially to classical conditioning procedures.
References
Allen, C.T., Janiszewski, C.A., 1989. Assessing the role of contingency awareness in
attitudinal conditioning with implications for advertising research. Journal of
Marketing Research 26 (February), 30–43.
Allen, C.T., Madden, T.J., 1985. A closer look at classical conditioning. Journal of
Consumer Research 12 (December), 301–315.
Baer, P.E., Fuhrer, M.J., 1982. Cognitive factors in the concurrent differential
conditioning of eyelid and skin conductance responses. Memory & Cognition 10
(March), 135–140.
Baeyens, F., Crombez, G., Van den Bergh, O., Eelen, P., 1988. Once in contact always
in contact: Evaluative conditioning is resistant to extinction. Advances in
Behavioral Research and Therapy 10, 179–199.
Baeyens, F., De Houwer, J., 1995. Evaluative conditioning is a qualitatively distinct
form of classical conditioning. Behaviour Research and Therapy 33, 825–831.
Baeyens, F., De Houwer, J., Vansteenwegen, D., Eelen, P., 1998. Evaluative
conditioning is a form of associative learning: On the artifactual nature of
Field and Davey’s (1997) artifactual account of evaluative learning. Learning
and Motivation 29, 461–474.
Baeyens, F., Eelen, P., Van den Bergh, O., 1990. Contingency awareness in evaluative
conditioning: A case of unaware affective-evaluative learning. Cognition and
Emotion 4, 3–18.
Baeyens, F., Hermans, D., Eelen, P., 1993. The role of CS–US contingency in human
evaluative conditioning. Behaviour Research and Therapy 31, 731–737.
Baker, W.E., 1999. When can affective conditioning and mere exposure directly
influence brand choice? Journal of Advertising 28, 31–46.
Barnes-Holmes, D., Keane, J., Barnes-Holmes, Y., Smeets, P.M., 2000. A derived
transfer of emotive functions as a means of establishing differential preferences
for soft drinks. Psychological Record 50 (3), 493–511.
Bierley, C., McSweeney, F.K., Vannieuwkerk, R., 1985. Classical conditioning of
preferences for stimuli. Journal of Consumer Research 12 (December), 316–323.
Biferno, M.A., Dawson, M.E., 1977. The onset of contingency awareness and
electrodermal classical conditioning: An analysis of temporal relationships
during acquisition and extinction. Psychophysiology 14 (March), 164–171.
Blair, M.E., Shimp, T.A., 1992. Consequences of an unpleasant experience with
music: A second-order negative conditioning perspective. Journal of Advertising
21 (1), 35–43.
Brown, P.L., Jenkins, H.M., 1968. Auto-shaping of the pigeon’s key-peck. Journal of
the Experimental Analysis of Behavior 11 (1), 1–8.
Cacioppo, J.T., Marshall-Goodell, B.S., Tassinary, L.G., Petty, R.E., 1992. Rudimentary
determinants of attitudes: Classical conditioning is more effective when prior
knowledge about the attitude stimulus is low than high. Journal of
Experimental Social Psychology 28 (May), 207–233.
Cosmopolitan, 2012. 2012 Display Advertising Rates. Available from: <http://
www.cosmomediakit.com/hotdata/publishers/cosmopoli2521681/categories/
Rates_Section.pdf>.
Crupi, A., 2011. In Their Prime: Broadcast Spot Costs Soar. Available from: <http://
www.adweek.com/news/television/their-prime-broadcast-spot-costs-soar132805>.
Dawson, M.E., 1973. Can classical conditioning occur without contingency learning?
A review and evaluation of the evidence. Psychophysiology 10 (January), 82–86.
De Houwer, J., Thomas, S., Baeyens, F., 2001. Associative learning likes and dislikes:
A review of 25 years of research on human evaluative conditioning.
Psychological Bulletin 127, 853–869.
Edell, J.A., Burke, M.C., 1984. The moderating effect of attitude toward an ad on ad
effectiveness under different processing conditions. In: Kinnear, T.C. (Ed.),
Advances in Consumer Research, vol. 11. Association for Consumer Research,
Provo, UT, pp. 644–649.
Feinberg, R., 1986. Credit cards as spending facilitating stimuli: A conditioning
interpretation. Journal of Consumer Research 13 (3), 348–356.
Field, A.P., 2000. I like it, but I’m not sure why: Can evaluative conditioning occur
without conscious awareness? Consciousness and Cognition 9, 13–36.
Field, A.P., Davey, G.C.L., 1999. Reevaluating evaluative conditioning: A
nonassociative explanation of conditioning effects in the visual evaluative
conditioning paradigm. Journal of Experimental Psychology: Animal Behavior
Processes 25, 211–224.
Garcia, J., Koelling, R.A., 1966. Relation of cue to consequence in avoidance learning.
Psychonomic Science 4 (3), 123–124.
295
Gibson, B., 2008. Can evaluative conditioning change attitudes toward mature
brands? New evidence from the implicit association test. Journal of Consumer
Research 35 (June), 178–188.
Gorn, G.J., 1982. The effects of music in advertising on choice behavior: A classical
conditioning approach. Journal of Marketing 46 (Winter), 94–101.
Greenwald, A.G., Leavitt, C., 1984. Audience involvement in advertising: Four levels.
Journal of Consumer Research 11 (June), 581–592.
Greenwald, A.G., McGhee, D.E., Schwartz, J.L.K., 1998. Measuring individual
differences in implicit cognition: The implicit association test. Journal of
Personality and Social Psychology 74 (June), 1464–1480.
Gresham, L.G., Shimp, T.A., 1985. Attitude toward the advertisement and brand
attitudes: A classical conditioning perspective. Journal of Advertising 14 (1), 10–
17, 49.
Groenland, E.A.G., Schoormans, J.P.L., 1994. Comparing mood-induction and
affective conditioning as mechanisms influencing product evaluation and
product choice. Psychology & Marketing 11, 183–197.
Grossman, R.P., Till, B.D., 1998. The persistence of classically conditioned brand
attitudes. Journal of Advertising 27 (1), 23–31.
Hammerl, M., 2000. I like it, but only when I’m not sure why: Evaluative
conditioning and the awareness issue. Consciousness and Cognition 9, 37–40.
Hammerl, M., Grabitz, H.J., 1996. Human evaluative conditioning without
experiencing a valued event. Learning and Motivation 27, 278–293.
Hunt, J.M., Florsheim, R.A., Chatterjee, A., Kernan, J.B., 1990. Credit cards as
spending-facilitating stimuli: A test and extension of Feinberg’s conditioning
hypothesis. Psychological Reports 67, 323–330.
Janiszewski, C., Warlop, L., 1993. The influence of classical conditioning procedures
on subsequent attention to the conditioned brand. Journal of Consumer
Research 20 (September), 171–189.
Kamin, L.J., 1969. Predictability, surprise, attention and conditioning. In: Campbell,
B.A., Church, R.M. (Eds.), Punishment and Aversive Behavior. Appleton-CenturyCrofts, New York, NY, pp. 279–296.
Kamin, L.J., Brimer, C.J., 1963. The effects of intensity of conditioned and
unconditioned stimuli on a conditioned emotional response. Canadian Journal
of Psychology 17 (2), 194–200.
Kassarjian, H.H., 1986. Consumer research: Some recollections and a commentary.
In: Lutz, R.J. (Ed.), Advances in Consumer Research, vol. 13. Association for
Consumer Research, Provo, UT, pp. 6–8.
Kellaris, J.J., Cox, A.D., 1989. The effects of background music in advertising: A
reassessment. Journal of Consumer Research 16 (June), 113–118.
Kim, J., Allen, C.T., Kardes, F.R., 1996. An investigation of the mediational
mechanisms underlying attitudinal conditioning. Journal of Marketing
Research 33 (August), 318–328.
Kim, J., Lim, J.S., Bhargava, M., 1998. The role of affect in attitude formation: A
classical conditioning approach. Journal of the Academy of Marketing Science
26 (2), 142–152.
Lie, C., Hunt, M., Peters, H.L., Veliu, B., Harper, D., 2010. The ‘negative’ credit card
effect: Credit cards as spending-limiting stimuli in New Zealand. Psychological
Record 60, 399–412.
Lovibond, P.F., Shanks, D.R., 2002. The role of awareness in Pavlovian conditioning:
Empirical evidence and theoretical implications. Journal of Experimental
Psychology: Animal Behavior Processes 28, 3–26.
Lubow, R.E., 1973. Latent inhibition. Psychological Bulletin 79 (6), 398–407.
MacKenzie, S.B., Lutz, R.J., Belch, G.E., 1986. The role of attitude toward the ad as a
mediator of advertising effectiveness: A test of competing explanations. Journal
of Marketing Research 23 (May), 130–143.
Macklin, M.C., 1986. Classical conditioning effects in product/character pairings
presented to children. In: Lutz, R.J. (Ed.), Advances in Consumer Research, vol.
13. Association for Consumer Research, Provo, UT, pp. 198–203.
Maus Media Group, 2011, March 8. How Much Does It Cost to Make a TV
Commercial? Available from: <http://mausmediagroup.wordpress.com/2011/
03/08/how-much-does-it-cost-to-make-a-tv-commercial>.
Martindale, C., 1991. Cognitive Psychology: A Neural-Network Approach. Brooks/
Cole, Pacific Grove, CA.
McCall, M., Belmont, H.J., 1996. Credit card insignia and restaurant tipping:
Evidence for an associative link. Journal of Applied Psychology 81, 609–613.
McCall, M., Trombetta, J., Gipe, A., 2004. Credit cues and impression management: A
preliminary attempt to explain the credit card effect. Psychological Reports 95,
331–337.
McSweeney, F.K., Bierley, C., 1984. Recent developments in classical conditioning.
Journal of Consumer Research 11 (September), 619–631.
Mis, F.W., Moore, J.W., 1973. Effect of preacquisition UCS stimulus exposure on
classical conditioning of the rabbit’s nictitating membrane response. Learning
and Motivation 4 (1), 108–114.
Monger, J.E., Feinberg, R.A., 1997. Mode of payment and formation of reference
prices. Pricing Strategy and Practice 5, 142–147.
Olson, M.A., Fazio, R.H., 2001. Implicit attitude formation through classical
conditioning. Psychological Science 12 (5), 413–417.
Pavlov, I.P., 1927. Conditioned Reflexes. Oxford University Press, Oxford, England.
Perruchet, P., 1985. Expectancy for airpuff and conditioned eyeblinks in humans.
Acta Psychologica 58 (January), 31–44.
Petty, R.E., Cacioppo, J.T., Schumann, D., 1983. Central and peripheral routes to
advertising effectiveness: The moderating role of involvement. Journal of
Consumer Research 10 (September), 135–146.
Prelec, D., Simester, D., 2001. Always leave home without it: A further investigation
of the credit-card effect on willingness to pay. Marketing Letters 12, 5–12.
296
C. Pornpitakpan / Australasian Marketing Journal 20 (2012) 282–296
Preston, I.L., 1982. The association model of the advertising communication process.
Journal of Advertising 11 (2), 3–15.
Priluck, R., Till, B.D., 2004. The role of contingency awareness, involvement, and
need for cognition in attitude formation. Journal of the Academy of Marketing
Science 32 (3), 329–404.
Rachman, S., 1991. Neo-conditioning and the classical theory of fear acquisition.
Clinical Psychology Review 11 (2), 155–173.
Raghubir, P., Srivastava, J., 2008. Monopoly money: The effect of payment coupling
and form on spending behavior. Journal of Experimental Psychology: Applied
14, 213–225.
Razran, G., 1938. Conditioning away social bias by the luncheon technique.
Psychological Bulletin 35, 693.
Rescorla, R.A., 1967. Pavlovian conditioning and its proper control procedures.
Psychological Review 74 (1), 71–80.
Rescorla, R.A., 1972. Informational variables in Pavlovian conditioning. The
Psychology of Learning and Motivation 6, 1–46.
Rescorla, R.A., 1973. Effect of US habituation following conditioning. Journal of
Comparative and Physiological Psychology 82 (1), 137–143.
Rescorla, R.A., 1988. Pavlovian conditioning: It’s not what you think it is. American
Psychologist 43 (March), 151–160.
Rescorla, R.A., Wagner, A.R., 1972. A theory of Pavlovian conditioning: Variations in
the effectiveness of reinforcement and nonreinforcement. In: Black, A.H.,
Prokasy, W.F. (Eds.), Classical Conditioning II. Appleton-Century-Crofts, New
York, pp. 64–99.
Rozin, P., Wrzesniewski, A., Byrnes, D., 1998. The elusiveness of evaluative
conditioning. Learning and Motivation 29, 397–415.
Sawyer, A.G., 1975. Demand artifacts in laboratory experiments in consumer
research. Journal of Consumer Research 1 (March), 20–29.
Schemer, C., Matthes, J., Wirth, W., Textor, S., 2008. Does ‘passing the courvoisier’
always pay off? Positive and negative evaluative conditioning effects of brand
placements in music videos. Psychology & Marketing 25 (10), 923–943.
Schiffman, L.G., Kanuk, L.L., 2010. Consumer Behavior, 10th ed. Prentice Hall, Upper
Saddle River, NJ.
Shimp, T.A., Moody, M.P., 2000. In search of a theoretical explanation for the credit
card effect. Journal of Business Research 48 (1), 17–23.
Shimp, T.A., Stuart, E.W., Engle, R.W., 1991. A program of classical conditioning
experiments testing variations in the conditioned stimulus and context. Journal
of Consumer Research 18 (June), 1–12.
Shurtleff, D., Ayres, J.J.B., 1981. One-trial backward excitatory fear conditioning in
rats: Acquisition, retention, extinction and spontaneous recovery. Animal
Learning and Behavior 9 (1), 65–74.
Spetch, M.L., Wilkie, D.M., Pinel, J.P., 1981. Backward conditioning: A reevaluation of
the empirical evidence. Psychological Bulletin 89 (January), 163–175.
Staats, A.W., Staats, C.K., 1959. Effect of number of trials on the language
conditioning of meaning. Journal of General Psychology 61 (October), 211–223.
Stahl, C., Unkelbach, C., Corneille, O., 2009. On the respective contributions of
awareness of unconditioned stimulus valence and unconditioned stimulus
identity in attitude formation through evaluative conditioning. Journal of
Personality and Social Psychology 97 (3), 404–420.
Stuart, E.W., Shimp, T.A., Engle, R.W., 1987. Classical conditioning of consumer
attitudes: Four experiments in an advertising context. Journal of Consumer
Research 14 (December), 334–349.
Terrace, H.S., Gibbon, J., Farrell, L., Baldock, M.D., 1975. Temporal factors influencing
the acquisition and maintenance of an autoshaped keypeck. Animal Learning
and Behavior 3 (1), 53–62.
Till, B.D., 1998. Using celebrity endorsers effectively: Lessons from associative
learning. Journal of Product and Brand Management 7 (5), 400–409.
Tom, G., 1995. Classical conditioning of unattended stimuli. Psychology &
Marketing 12, 79–87.
Wagner, A.R., Siegel, S., Thomas, E., Ellison, G.D., 1964. Reinforcement history and
the extinction of a conditioned salivary response. Journal of Comparative and
Physiological Psychology 58 (3), 354–358.
Walther, E., 2002. Guilty by mere association: Evaluative conditioning and the
spreading attitude effect. Journal of Personality and Social Psychology 82, 919–
934.
Walther, E., Grigoriadis, S., 2004. Why sad people like shoes better: The influence of
mood on the evaluative conditioning of consumer attitudes. Psychology &
Marketing 21, 75–773.
Zajonc, R.R., 1968. Attitudinal effects of mere exposure. Journal of Personality and
Social Psychology Monograph Supplement 9 (Part 2, June), 1–27.